Application and comparison of classification algorithms for recognition of Alzheimer's disease in electrical brain activity (EEG)

The early detection of subjects with probable Alzheimer's disease (AD) is crucial for effective appliance of treatment strategies. Here we explored the ability of a multitude of linear and non-linear classification algorithms to discriminate between the electroencephalograms (EEGs) of patients with varying degree of AD and their age-matched control subjects. Absolute and relative spectral power, distribution of spectral power, and measures of spatial synchronization were calculated from recordings of resting eyes-closed continuous EEGs of 45 healthy controls, 116 patients with mild AD and 81 patients with moderate AD, recruited in two different centers (Stockholm, New York). The applied classification algorithms were: principal component linear discriminant analysis (PC LDA), partial least squares LDA (PLS LDA), principal component logistic regression (PC LR), partial least squares logistic regression (PLS LR), bagging, random forest, support vector machines (SVM) and feed-forward neural network. Based on 10-fold cross-validation runs it could be demonstrated that even tough modern computer-intensive classification algorithms such as random forests, SVM and neural networks show a slight superiority, more classical classification algorithms performed nearly equally well. Using random forests classification a considerable sensitivity of up to 85% and a specificity of 78%, respectively for the test of even only mild AD patients has been reached, whereas for the comparison of moderate AD vs. controls, using SVM and neural networks, values of 89% and 88% for sensitivity and specificity were achieved. Such a remarkable performance proves the value of these classification algorithms for clinical diagnostics.

Classification Using Generalized Partial Least Squares

The advances in computational biology have made simultaneous monitoring of thousands of features possible. The high throughput technologies not only bring about a much richer information context in which to study various aspects of gene functions but they also present challenge of analyzing data with large number of covariates and few samples. As an integral part of machine learning, classification of samples into two or more categories is almost always of interest to scientists. In this paper, we address the question of classification in this setting by extending partial least squares (PLS), a popular dimension reduction tool in chemometrics, in the context of generalized linear regression based on a previous approach, Iteratively ReWeighted Partial Least Squares, i.e. IRWPLS (Marx, 1996). We compare our results with two-stage PLS (Nguyen and Rocke, 2002A; Nguyen and Rocke, 2002B) and other classifiers. We show that by phrasing the problem in a generalized linear model setting and by applying bias correction to the likelihood to avoid (quasi)separation, we often get lower classification error rates.

Model Evaluation Based on the Distribution of Estimated Absolute Prediction Error

The construction of a reliable, practically useful prediction rule for future response is heavily dependent on the "adequacy" of the fitted regression model. In this article, we consider the absolute prediction error, the expected value of the absolute difference between the future and predicted responses, as the model evaluation criterion. This prediction error is easier to interpret than the average squared error and is equivalent to the mis-classification error for the binary outcome. We show that the distributions of the apparent error and its cross-validation counterparts are approximately normal even under a misspecified fitted model. When the prediction rule is "unsmooth", the variance of the above normal distribution can be estimated well via a perturbation-resampling method. We also show how to approximate the distribution of the difference of the estimated prediction errors from two competing models. With two real examples, we demonstrate that the resulting interval estimates for prediction errors provide much more information about model adequacy than the point estimates alone.

Evaluating Prediction Rules for t-Year Survivors With Censored Regression Models

Suppose that we are interested in establishing simple, but reliable rules for predicting future t-year survivors via censored regression models. In this article, we present inference procedures for evaluating such binary classification rules based on various prediction precision measures quantified by the overall misclassification rate, sensitivity and specificity, and positive and negative predictive values. Specifically, under various working models we derive consistent estimators for the above measures via substitution and cross validation estimation procedures. Furthermore, we provide large sample approximations to the distributions of these nonsmooth estimators without assuming that the working model is correctly specified. Confidence intervals, for example, for the difference of the precision measures between two competing rules can then be constructed. All the proposals are illustrated with two real examples and their finite sample properties are evaluated via a simulation study.

A coronary heart disease risk model for predicting the effect of potent antiretroviral therapy in HIV-1 infected men

BACKGROUND: Many HIV-infected patients on highly active antiretroviral therapy (HAART) experience metabolic complications including dyslipidaemia and insulin resistance, which may increase their coronary heart disease (CHD) risk. We developed a prognostic model for CHD tailored to the changes in risk factors observed in patients starting HAART. METHODS: Data from five cohort studies (British Regional Heart Study, Caerphilly and Speedwell Studies, Framingham Offspring Study, Whitehall II) on 13,100 men aged 40-70 and 114,443 years of follow up were used. CHD was defined as myocardial infarction or death from CHD. Model fit was assessed using the Akaike Information Criterion; generalizability across cohorts was examined using internal-external cross-validation. RESULTS: A parametric model based on the Gompertz distribution generalized best. Variables included in the model were systolic blood pressure, total cholesterol, high-density lipoprotein cholesterol, triglyceride, glucose, diabetes mellitus, body mass index and smoking status. Compared with patients not on HAART, the estimated CHD hazard ratio (HR) for patients on HAART was 1.46 (95% CI 1.15-1.86) for moderate and 2.48 (95% CI 1.76-3.51) for severe metabolic complications. CONCLUSIONS: The change in the risk of CHD in HIV-infected men starting HAART can be estimated based on typical changes in risk factors, assuming that HRs estimated using data from non-infected men are applicable to HIV-infected men. Based on this model the risk of CHD is likely to increase, but increases may often be modest, and could be offset by lifestyle changes.

Oral purified bacterial extracts in acute respiratory tract infections in childhood: a systematic quantitative review

BACKGROUND: Recurrent acute respiratory tract infections (ARTI) are a common problem in childhood. Some evidence suggests a benefit regarding the prevention of ARTI in children treated with the immunomodulator OM-85 BV (Bronchovaxom). METHODS: We summarised the evidence on the effectiveness of the immunomodulator OM-85 BV in the prevention of ARTI in children. We searched randomised comparisons of oral purified bacterial extracts against inactive controls in children with respiratory tract diseases in nine electronic databases and reference lists of included studies. We extracted salient features of each study, calculated relative risks (RR) or weighted mean differences (WMD) and performed meta-analyses using random-effects models. RESULTS: Thirteen studies (2,721 patients) of low to moderate quality tested OM-85 BV. Patients and outcomes differed substantially, which impeded pooling results of more than two trials. Two studies (240 patients) reporting on the number of patients with less than three infections over 6 month of follow-up in children not in day care showed a trend for benefit RR 0.82 (95% CI, 0.65-1.02). One out of two studies examining the number of children not in day care without infections over 4-6 month reported a significant RR of 0.42 (95% CI, 0.21-0.82) whereas the smaller, second study did not [RR 0.92 (95% CI, 0.58-1.46)]. Two studies reporting the number of antibiotic courses indicated a benefit for the intervention arm [WMD 2.0 (95% CI, 1.7-2.3)]. Two out of the three studies showed a reduction of length of episodes of 4-6 days whereas a third study showed no difference between the two groups. CONCLUSION: Evidence in favour of OM-85 BV in the prevention of ARTI in children is weak. There is a trend for fewer and shorter infections and a reduction of antibiotic use.

An efficient framework for photon Monte Carlo treatment planning

Currently photon Monte Carlo treatment planning (MCTP) for a patient stored in the patient database of a treatment planning system (TPS) can usually only be performed using a cumbersome multi-step procedure where many user interactions are needed. This means automation is needed for usage in clinical routine. In addition, because of the long computing time in MCTP, optimization of the MC calculations is essential. For these purposes a new graphical user interface (GUI)-based photon MC environment has been developed resulting in a very flexible framework. By this means appropriate MC transport methods are assigned to different geometric regions by still benefiting from the features included in the TPS. In order to provide a flexible MC environment, the MC particle transport has been divided into different parts: the source, beam modifiers and the patient. The source part includes the phase-space source, source models and full MC transport through the treatment head. The beam modifier part consists of one module for each beam modifier. To simulate the radiation transport through each individual beam modifier, one out of three full MC transport codes can be selected independently. Additionally, for each beam modifier a simple or an exact geometry can be chosen. Thereby, different complexity levels of radiation transport are applied during the simulation. For the patient dose calculation, two different MC codes are available. A special plug-in in Eclipse providing all necessary information by means of Dicom streams was used to start the developed MC GUI. The implementation of this framework separates the MC transport from the geometry and the modules pass the particles in memory; hence, no files are used as the interface. The implementation is realized for 6 and 15 MV beams of a Varian Clinac 2300 C/D. Several applications demonstrate the usefulness of the framework. Apart from applications dealing with the beam modifiers, two patient cases are shown. Thereby, comparisons are performed between MC calculated dose distributions and those calculated by a pencil beam or the AAA algorithm. Interfacing this flexible and efficient MC environment with Eclipse allows a widespread use for all kinds of investigations from timing and benchmarking studies to clinical patient studies. Additionally, it is possible to add modules keeping the system highly flexible and efficient.

Practice and perception--a nationwide survey of therapy habits in sepsis

OBJECTIVE: To simultaneously determine perceived vs. practiced adherence to recommended interventions for the treatment of severe sepsis or septic shock. DESIGN: One-day cross-sectional survey. SETTING: Representative sample of German intensive care units stratified by hospital size. PATIENTS: Adult patients with severe sepsis or septic shock. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Practice recommendations were selected by German Sepsis Competence Network (SepNet) investigators. External intensivists visited intensive care units randomly chosen and asked the responsible intensive care unit director how often these recommendations were used. Responses "always" and "frequently" were combined to depict perceived adherence. Thereafter patient files were audited. Three hundred sixty-six patients on 214 intensive care units fulfilled the criteria and received full support. One hundred fifty-two patients had acute lung injury or acute respiratory distress syndrome. Low-tidal volume ventilation < or = 6 mL/kg/predicted body weight was documented in 2.6% of these patients. A total of 17.1% patients had tidal volume between 6 and 8 mL/kg predicted body weight and 80.3% > 8 mL/kg predicted body weight. Mean tidal volume was 10.0 +/- 2.4 mL/kg predicted body weight. Perceived adherence to low-tidal volume ventilation was 79.9%. Euglycemia (4.4-6.1 mmol/L) was documented in 6.2% of 355 patients. A total of 33.8% of patients had blood glucose levels < or = 8.3 mmol/L and 66.2% were hyperglycemic (blood glucose > 8.3 mmol/L). Among 207 patients receiving insulin therapy, 1.9% were euglycemic, 20.8% had blood glucose levels < or = 8.3 mmol/L, and 1.0% were hypoglycemic. Overall, mean maximal glucose level was 10.0 +/- 3.6 mmol/L. Perceived adherence to strict glycemic control was 65.9%. Although perceived adherence to recommendations was higher in academic and larger hospitals, actual practice was not significantly influenced by hospital size or university affiliation. CONCLUSIONS: This representative survey shows that current therapy of severe sepsis in German intensive care units complies poorly with practice recommendations. Intensive care unit directors perceive adherence to be higher than it actually is. Implementation strategies involving all intensive care unit staff are needed to overcome this gap between current evidence-based knowledge, practice, and perception.

Use of hydroclimatic forecasts for improved water management in central Texas

Accurate seasonal to interannual streamflow forecasts based on climate information are critical for optimal management and operation of water resources systems. Considering most water supply systems are multipurpose, operating these systems to meet increasing demand under the growing stresses of climate variability and climate change, population and economic growth, and environmental concerns could be very challenging. This study was to investigate improvement in water resources systems management through the use of seasonal climate forecasts. Hydrological persistence (streamflow and precipitation) and large-scale recurrent oceanic-atmospheric patterns such as the El Niño/Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), North Atlantic Oscillation (NAO), the Atlantic Multidecadal Oscillation (AMO), the Pacific North American (PNA), and customized sea surface temperature (SST) indices were investigated for their potential to improve streamflow forecast accuracy and increase forecast lead-time in a river basin in central Texas. First, an ordinal polytomous logistic regression approach is proposed as a means of incorporating multiple predictor variables into a probabilistic forecast model. Forecast performance is assessed through a cross-validation procedure, using distributions-oriented metrics, and implications for decision making are discussed. Results indicate that, of the predictors evaluated, only hydrologic persistence and Pacific Ocean sea surface temperature patterns associated with ENSO and PDO provide forecasts which are statistically better than climatology. Secondly, a class of data mining techniques, known as tree-structured models, is investigated to address the nonlinear dynamics of climate teleconnections and screen promising probabilistic streamflow forecast models for river-reservoir systems. Results show that the tree-structured models can effectively capture the nonlinear features hidden in the data. Skill scores of probabilistic forecasts generated by both classification trees and logistic regression trees indicate that seasonal inflows throughout the system can be predicted with sufficient accuracy to improve water management, especially in the winter and spring seasons in central Texas. Lastly, a simplified two-stage stochastic economic-optimization model was proposed to investigate improvement in water use efficiency and the potential value of using seasonal forecasts, under the assumption of optimal decision making under uncertainty. Model results demonstrate that incorporating the probabilistic inflow forecasts into the optimization model can provide a significant improvement in seasonal water contract benefits over climatology, with lower average deficits (increased reliability) for a given average contract amount, or improved mean contract benefits for a given level of reliability compared to climatology. The results also illustrate the trade-off between the expected contract amount and reliability, i.e., larger contracts can be signed at greater risk.

Extracorporeal membrane oxygenation (ECMO): extended indications for artificial support of both heart and lungs

Extracorporeal membrane oxygenation (ECMO) was used to achieve temporary artificial support in cardiac and pulmonary function in 22 patients from 1987 to September 1990. Standard indications were postcardiotomy cardiogenic shock (n = 4), neonatal (n = 1) and adult respiratory distress syndrome (n = 4). ECMO was also used for extended indications, such as graft failure following heart (n = 11) or lung transplantation (n = 2). In six of these cases ECMO was instituted as a bridge device to subsequent retransplantation of either the heart (n = 4) or one lung (n = 2). One out of nine patients supported by ECMO for standard indications, and two out of 13 patients supported for extended indications are long-term survivors. This series illustrates the results with ECMO in emergency situations, in patients under immunosuppressive protocols, or in patients with advanced lung failure requiring almost complete artificial gas exchange. In such complex situations, ECMO does provide stabilization until additional therapeutic measures are in effect. ECMO cannot be recommended for postoperative cardiogenic shock but short-term ECMO support is an accepted method in most cases with graft failure or pulmonary failure or other origin.

[New insights into diagnosis and management of gestational diabetes mellitus: recommendations of the Swiss Society for Endocrinology and Diabetes]

Physiology and current knowledge about gestational diabetes which led to the adoption of new diagnostic criterias and blood glucose target levels during pregnancy by the Swiss Society for Endocrinology and Diabetes are reviewed. The 6th International Workshop Conference on Gestational Diabetes mellitus in Pasedena (2008) defined new diagnostic criteria based on the results of the HAPO-Trial. These criteria were during the ADA congress in New Orleans in 2009 presented. According to the new criteria there is no need for screening, but all pregnant women have to be tested with a 75 g oral glucose tolerance test between the 24th and 28th week of pregnancy. The new diagnostic values are very similar to the ones previously adopted by the ADA with the exception that only one out of three values has to be elevated in order to make the diagnosis of gestational diabetes. Due to this important difference it is very likely that gestational diabetes will be diagnosed more frequently in the future. The diagnostic criteria are: Fasting plasma glucose > or = 5.1 mmol/l, 1-hour value > or = 10.0 mmol/l or 2-hour value > or = 8.5 mmol/l. Based on current knowledge and randomized trials it is much more difficult to define glucose target levels during pregnancy. This difficulty has led to many different recommendations issued by diabetes societies. The Swiss Society of Endocrinology and Diabetes follows the arguments of the International Diabetes Federation (IDF) that self-blood glucose monitoring itself lacks precision and that there are very few randomized trials. Therefore, the target levels have to be easy to remember and might be slightly different in mmol/l or mg/dl. The Swiss Society for Endocrinology and Diabetes adopts the tentative target values of the IDF with fasting plasma glucose values < 5.3 mM and 1- and 2-hour postprandial (after the end of the meal) values of < 8.0 and 7.0 mmol/l, respectively. The last part of these recommendations deals with the therapeutic options during pregnancy (nutrition, physical exercise and pharmaceutical treatment). If despite lifestyle changes the target values are not met, approximately 25 % of patients have to be treated pharmaceutically. Insulin therapy is still the preferred treatment option, but metformin (and as an exception glibenclamide) can be used, if there are major hurdles for the initiation of insulin therapy.